Who we work with/Law firms

Law firms

AI agent development for law firms. Control first, novelty second.

Agents that start on their own when a new matter opens, a meeting ends or post arrives, do the checking, drafting and filing, then stop for the fee earner to approve. Built in your own Microsoft 365.

Book a free 30-minute call See the agents

FiveForward builds AI agents for UK law firms in Copilot Studio and Power Automate, inside the firm's own Microsoft 365: new matter intake and conflict checks, attendance notes, post and email routed to the right matter and first drafts of client care letters, with the fee earner approving before anything goes out.

Law firms are right to be cautious about AI. Agents sit close to matter files, client emails and privileged material, so building them takes more than enthusiasm. Control first, novelty second: every agent has one job, a defined trigger and a point where the fee earner approves.

The agents that give a firm time back are the ones that run. They start when a new matter is opened, a meeting ends or a letter arrives, do the checking, drafting and filing in your own Microsoft 365, then stop. Nothing goes to a client and nothing is filed as final until a person has approved it.

Advice, verification and sign-off stay with people. The agent prepares; the fee earner decides.

Run an accountancy firm? The accountancy page has the detail: six ready-made agents, what starts each one and where your team approves.

The agents

Four agents that run in a law firm.

Each one starts on its own, does the preparation and stops for the fee earner.

New matter intake and conflict checks

Starts when
A new matter or enquiry is opened.
What it does
Gathers the client and matter details, pulls Companies House data, runs the conflict search across your clients and matters plus your approved checks, and files the report on the matter with an evidence line.
Where a person approves
The fee earner or compliance reviews any hits before the matter is accepted.

Attendance notes

Starts when
A client meeting ends.
What it does
Drafts the attendance note in your template from the transcript, with the actions, dates and any changed instructions, and puts a follow-up email in drafts.
Where a person approves
The fee earner approves or amends the note before it is filed to the matter.
Runs on a ready-made agent

Post and email to the right matter

Starts when
Post is scanned or an email reaches a shared inbox.
What it does
Reads each item, matches it to the client and matter, files it with your naming and tells the right person in Teams.
Where a person approves
Anything it can't match with confidence goes to a person, not a guess.

First drafts of client care letters

Starts when
A matter is accepted.
What it does
Assembles the client care letter from your approved templates, with the right fee earner, scope and fees for the matter type.
Where a person approves
The fee earner reviews and signs, and nothing is sent until they have.

Proof

The same pattern, from accountancy.

These come from accountancy practice, and the sector label on each card is the real one, so you can see exactly where the parallel holds.

Worked example · Accountancy

How a UK accountancy firm built an AI agent that reads, understands and files the post

Built for an accountancy firm: scanned post read, matched against more than 1,500 live clients and filed in about two minutes, with exceptions sent to a person. The same pattern routes post and email to the right matter in a law firm.

Worked example · Client acceptance

Collapsing letter of engagement drafting from half an hour to a few minutes of form filling

Letters of engagement generated from a form into controlled Word templates, so the partner reviews rather than drafts. The same pattern produces client care letters and standard matter correspondence.

What to watch

Where agents go wrong, and how we avoid it.

  • Matter permissions that were already too broad, so an agent people talk to surfaces content that should have been locked down first
  • An agent working from an untidy precedent library and confidently citing superseded wording
  • Privileged or confidential material pasted into public AI tools because the firm has no governed alternative of its own
  • An agent's summary of a matter file treated as the source document, so the detail that mattered never gets read

Your data and who approves

Built in your Microsoft 365. Approved by your team.

You decide what each agent may do on its own and where a person has to approve. By default, anything that goes to a client or raises a bill waits for one of your team.

Built in your Microsoft 365

Agents are built in your own Microsoft 365, and your own Azure subscription where a job needs it, not on our servers. We don't host your data.

Agents you talk to see what you can see

An agent your team talks to only sees what that person can already see.

Agents that run use a service account

Agents that run on their own use a dedicated service account. Its access is limited to the mailboxes, folders and systems the job needs, agreed with you in advance.

Every run is logged

Every run is logged: what came in, what the agent did and who approved it.

What it costs

Fixed prices, agreed before anything is built.

See pricing
Discover
£1,500 fixed, covering Design, credited back once follow-on spend passes £3,500
A build
From £5,000, typically £5,000 to £15,000, fixed at Design
Once live
Embed from £750 a month

Microsoft's own usage charges, such as Copilot Credits, are billed by Microsoft to your firm, separately from our fees. All prices exclude VAT.

Is our client data safe with AI agents?
We build agents in your own Microsoft 365, not on our servers, and we don't host your data. Agents your team talks to only see what that person can already see. Agents that run on their own use a dedicated service account, limited to the mailboxes, folders and systems the job needs, agreed with you in advance. We build in UK regions wherever Microsoft offers them, and anything that has to run elsewhere is set out for your compliance lead to approve before go-live. Microsoft doesn't use your Microsoft 365 content to train its foundation models, and every run is logged. For agents fee earners talk to, matter permissions and information barriers keep doing their job, which is why Discover checks them before anything is built.
How do agents handle privilege and confidentiality?
An agent a fee earner talks to only surfaces what that person could already open, so the protection comes from matter permissions and information barriers being correct first. Agents that run on their own use a service account limited to the named libraries and mailboxes their job needs. We treat the permissions tidy-up as part of Discover, not an afterthought.
Can an AI agent draft legal advice?
No agent we build is treated as a source of advice. Agents draft notes, letters and summaries from approved material, and the output is preparation for a qualified person. Legal judgement, verification and final responsibility stay with the lawyer.
Who approves what the agent does?
You decide what each agent may do on its own and where a person has to approve. By default, anything that goes to a client or raises a bill waits for one of your team. In a law firm that usually means the fee earner on the matter.
What do law firms usually build first?
New matter intake and conflict checks, attendance notes and post routing. Each has a clear trigger, an owner and a point where a fee earner approves. Attendance notes run on the ready-made client meeting agent; the others are built from your Discover plan.
How long does it take?
Most builds take two to six weeks from sign-off, depending on the systems involved.
What does AI agent development cost for a law firm?
The first step is a free 30-minute call. Discover is £1,500 fixed, covering Design, and credited back in full once follow-on spend passes £3,500 before VAT within 90 days of the roadmap. Builds are from £5,000, typically £5,000 to £15,000, fixed at Design before anything is built. Once live, Embed looks after your agents from £750 a month, and Microsoft bills its own usage to your firm.
How do we get started?
Book a free 30-minute call. No pitch and no obligation. Most firms start with Discover. If you already know what you want, such as one of the six accountancy agents, we scope it on the free call and the specification is included in the fixed build price. Discover ends with a plan on a page, played back in about an hour, where you choose the route and the price is fixed.

Further reading

Where an agent fits in a 20-person law firm

Intake, correspondence, conflict search and chronologies for an agent; advice and judgement for the fee earner.

Choosing a Copilot model for legal document review

Which model suits long legal documents, and where a person still has to read the source.

The client meeting agent

The ready-made agent behind attendance notes: a brief before the meeting, a note in your template after it.

Other sectors

AI agents for accountancy firms

Six ready-made agents for the admin every practice shares: WIP and billing, client checks, meeting notes, records chasing, onboarding and VAT review. Each one waits for one of your team to approve.

AI agents for insurance brokers

Agents that start on their own ahead of a renewal date, when a risk is ready for market or a claim goes quiet, do the preparation and chasing, then stop for the broker to approve. Built in your own Microsoft 365.

AI agents for financial advisers

Agents that brief the adviser before a review, draft the file note and a first suitability letter afterwards and pull provider responses into your schedule, then stop for the adviser to approve. Built in your own Microsoft 365.

AI agents for professional services firms

Agents that start on their own when an email arrives, a meeting ends or a report is exported, do the checking, drafting and filing, then stop for one of your team to approve.

Next step

Talk through the first agent your firm would hand work to.

A free 30-minute call with James. No pitch and no obligation.

Book a free 30-minute call Who we work with